Enhancement of Localization Performance in Binaural Synthesis

Head-related transfer functions (HRTFs) are the core of binaural synthesis as they provide the necessary cues to perceive spatial acoustics of virtual environments. Even though the use of generic HRTFs (listening through other's ears) represents the state-of-the-art in binaural synthesis, they proved lower performance in localization tasks than individual HRTFs (listening through our own ears).

The current project emerged as a proposal to explore the nature of the deteriorated performance in localization when a limited number of generic HRTFs are used in binaural synthesis. It seeks to bring together insights from physics and psychoacoustics to provide a better account of the perception of sound in virtual environments. The findings are intended to enhance binaural synthesis in terms of computational requirements and performance.

The goal of the project is to identify those HRTF features that enable localization. This parameterization will allow achieving different objectives of practical importance in the implementation of binaural synthesis: to optimize the number of discrete points for which HRTFs are needed in order to synthesize binaural sound in a full sphere, to obtain a data set of generic HRTFs with improved performance in localization tasks for binaural synthesis and to minimize the computational cost required in handling this data set.